You've to sample at the receiver with dopple sample frequency, when you
obtain your real IFFT output from a conjugate extended input-vector.
So you have 4 out of 8 samples again. Calculating the FFT yields your
complex data values (assuming ideal channel conditions).

see ya

Steffen

PS: try this "Multi-Carrier Digital Communications - Theory and
Applications for OFDM" - Author: A.Bahai, B.Saltzberg ISBN:0-306-46296-6

		
This message was sent using the Comp.DSP web interface on
www.DSPRelated.com

I have a conceptual doubt related to the issue mentioned below.The output
signal from a OFDM transmitter is the low pass filtered version of the
real part of IDFT operation.Now suppose,I have 4 complex data per
symbol,there fore,its IFFT operation yields me 4 sample point.
                                                             
theoretically,I obtain the real part of a DFT output by making conjugate
symmetry and then taking IFFT.Now this means that I have a 8 point
output-

example, suppse i have following data vector x=[a b c d];
now according to the theory of OFDM, my output samples in time domain
are-
Re{FFT(x)}.....this yields me 4 output samples.
Now to achieve the same in simulation and in real life implementation, we
do the following
we take an IFFT of the following series to get a real out put-
x1=[0 a b c d d* c* b* a* 0]
this gives me a real series...
                              Now my question is that, how do we justify
the difference between the output of Re(FFT(x)) and the ifft of x1...i.e
what is the reason for increased number of sample points...and second, the
values are also different...how do we justify that.
Looking for a prompt reply

thanks and regards
nishchal

>Dan,
>
>I had been playing around with OFDM generation recently. 
>Let me try to explain what could be done.
>
>Let [a,b,c,d] be your 4 cmplex message points.
>N= the size of the fft. 
>Place the 4 complex points as follows (in frequency)
>For odd N=9
>FFT is Xf(k) = [0 a b c d d* c* b* a* ]; k=0 to 8 (i.e. N-1)
>[* implies complex conjugation]
>The k=0 entry, the DC term is set to zero.
>An inverse-fft of Xf will give the (real) OFDM signal, x(n),say.
>If you choose N to be even, leave the Xf(k)=0, for k=0 and k=N/2;
>
>To get back the message points from the OFDM signal, find fft of x(n)
>
>Hope this sort few of your questions.
>
>Parthasarathy.
>




		
This message was sent using the Comp.DSP web interface on
www.DSPRelated.com

Dan,

I had been playing around with OFDM generation recently. 
Let me try to explain what could be done.

Let [a,b,c,d] be your 4 cmplex message points.
N= the size of the fft. 
Place the 4 complex points as follows (in frequency)
For odd N=9
FFT is Xf(k) = [0 a b c d d* c* b* a* ]; k=0 to 8 (i.e. N-1)
[* implies complex conjugation]
The k=0 entry, the DC term is set to zero.
An inverse-fft of Xf will give the (real) OFDM signal, x(n),say.
If you choose N to be even, leave the Xf(k)=0, for k=0 and k=N/2;

To get back the message points from the OFDM signal, find fft of x(n)

Hope this sort few of your questions.

Parthasarathy.

I am attempting to simulate a baseband OFDM communication system in
MATLAB.  I have the parallelized QAM symbols feeding into an IFFT, but
this produces a vector of complex numbers.

I found much information about taking the real and imaginary parts and
modulating them with an In-phase and Quadrature signal, however this
is not what I am trying to do.  I am simulating OFDM in the base band,
so I can not do this.  Instead, I need to produce a real sequence from
the IFFT so I can send it directly to a D/A.

I know that, in order to produce a real sequence, the inputs to the
IFFT must be conjugate-symmetric.  I am still having trouble, however.

Suppose, for simplicity, my system is to transmit 4 complex QAM
symbols at a time.  How do I arrange the 4-element QAM vector to
produce a real sequence via IFFT?  Do these 4 elements correspond to
DC and 3 positive frequencies?  If so, what do I need to add?  3
elements at the beginning of the vector corresponding to the
conjugates of the 3 positive frequencies?  If so, this would lead to a
seven point IFFT, which would produce a vector of 7 real numbers. 
Does this mean that I have to transmit on 7 subcarriers in order to
recover my 4 symbols? How do I extract my original 4 QAM symbols from
this? Simply an FFT?

Thanks in advance,

Dan Savio